Reporting an executive can lean on.

Why dashboards contradict themselves

Not because anyone lies — because different widgets are computed by different code from different queries at different times. One card counts drafts, another doesn't; one parser reads "1,500,000" as a number, another silently drops it; one total is scoped to your permissions, the total beside it isn't. Each choice is defensible alone. Together they produce a page that argues with itself.

The standard

One computation, every surface. The dashboard, the executive brief on screen, and the downloaded brief render from the same functions — identical numbers by construction, not by discipline.

Every figure says what it counts.Populations and denominators are named ("of work with a type recorded", "requests only"), and money is never rounded into a lie — $2.35M stays $2.35M.

Hidden is disclosed, never silent.When permissions or privacy floors keep records out of a view, the page says so — and says the totals understate. A figure that can't be read is omitted, never shown as a fabricated zero.

Every gap carries its explanation. Released exceeding committed names the cancelled work behind it. Committed exceeding budget turns red — the one red rule. Money on hold says so, in dollars.

Checked by a machine, on every release

Vision runs a coherence validator in its build pipeline: thousands of randomized portfolios per release, asserting that every pair of figures that should agree provably does — counts partition, money cross-foots to the dollar, and the insight sentences recompute from the same rollups the tables render. If any two numbers could disagree, the release does not ship. We believe no other portfolio product makes that guarantee.

Investment by strategic objective — money counted once

← Back to the practice guide